Pedestrian Planet: What YouTube Driving from 233 Countries and Territories Teaches Us About the World

Alam, Md Shadab; Martens, Marieke H.; Bazilinskyy, Pavlo · 2025 · OpenAlex-citations

DOI: 10.1145/3744333.3747827

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study addresses the critical need for global, cross-cultural data on pedestrian behavior to improve road safety and develop robust algorithms for automated vehicles (AVs) and advanced driver assistance systems (ADAS). Current research is often limited by small sample sizes, controlled environments, or single-city studies, which restricts the generalizability of pedestrian detection models. Motivated by the high global burden of traffic fatalities among vulnerable road users, the authors aim to provide a comprehensive analysis of pedestrian crossing behavior across diverse socioeconomic and infrastructural contexts. The researchers utilized the City Road Observations With Dashcams (CROWD) dataset, comprising dashcam footage sourced from YouTube. The study analyzed 8,494 videos totaling over 4,100 hours of footage from 233 countries and territories. Using the YOLOv11x object detection algorithm and ByteTrack tracker, the team identified and tracked pedestrians. They defined crossing events based on lateral trajectory movement and applied rigorous filters to exclude riders, camera-induced false positives, and implausible speeds. Key metrics measured were crossing initiation time (the delay between intent and stepping onto the road) and crossing speed. These behavioral metrics were correlated with contextual variables, including traffic mortality rates, GDP, Gini coefficient, literacy rates, and median age. The analysis revealed significant global variations in pedestrian behavior. The worldwide average crossing initiation time was 3.18 seconds, and the average crossing speed was 1.20 m/s. Qatar exhibited the longest mean initiation time (6.44 s), while China showed the fastest crossing speed (1.69 m/s). The study found a negative correlation between initiation time and crossing speed ($r = -0.18$), suggesting that pedestrians who hesitate longer tend to cross more slowly. Furthermore, crossing speed was negatively correlated with the Gini coefficient ($r = -0.19$) and positively correlated with traffic mortality ($r = 0.18$). Notably, countries with vastly different infrastructures, such as Bangladesh and the Netherlands, displayed similar crossing initiation times, highlighting complex behavioral adaptations. These findings underscore the importance of culturally aware road design and the development of adaptive interfaces for AVs. The study demonstrates that large-scale, real-world video data can effectively capture the diversity of global urban traffic, providing essential insights for creating scalable, context-aware safety systems. By linking pedestrian behavior to socioeconomic indicators, the research offers a framework for understanding how infrastructure and cultural norms influence road safety, ultimately supporting the creation of interventions that reduce accidents and save lives globally.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• cultural cross national